Senior leaders already sit through briefings on the tools. The gap is the next decision they have to sign: which workflow changes, which risk they will accept, and what they will tell their own team is allowed. A session that only names products leaves that signature where it was.
PwC put a leadership track on the same syllabus as the prompting. In How PwC is using generative AI to deliver business value, the firm describes a year into a $1 billion US investment. It had become OpenAI's first reseller of ChatGPT Enterprise and the largest user of the product, and it had equipped a 75,000-person US workforce. The upskilling programme, My AI, covers Responsible AI, generative AI prompting, and leadership in the age of generative AI. PwC says 95 percent of US employees took part and put in more than 360,000 hours. People who use the tools regularly show productivity gains of 20 to 40 percent. The same account cites PwC's Global CEO Survey: 68 percent of CEOs expect generative AI to increase employee efficiency this year, and 44 percent expect it to increase profits.
What the executive session has to teach
PwC's guideline on the work is specific. A company on vendor models upskills the people it already has, and it redesigns the work around the tools. The larger gain, in their telling, shows up when the tool takes simple tasks and people move to higher-value work. Responsible AI, as they describe it, keeps well-informed people in control of high-risk and high-value decisions.
That is the syllabus a leadership team should be able to walk out with. They should be able to look at a proposed use and say what decision it changes, what data it touches, and who still signs the output. They should be able to run one task from their own week on data they are allowed to use, so the standard is visible to the people who report to them. A leader who has never produced a page with the tool will struggle to judge the pages coming back from the team. The PwC split is a useful checklist for the brief you send a provider: responsible use, hands-on work, and a leadership track that is actually about decisions.
The survey numbers belong in that brief. Most chief executives in it expect efficiency. A smaller share expect profit. A faster draft of the old process can produce the first and miss the second. If the day never asks which meeting, report or approval will run differently next month, the room leaves with a stronger vocabulary and the same calendar. AI training for executives earns the date when it is built around that change.
Pick the artefact before you call
PwC's own examples are tied to the work of a function. Finance teams draft contracts and pull terms out of existing ones. Software teams review code and write documentation. Marketing teams produce content. Your executive group needs the equivalent object from this company: a board risk note, a customer proposal, a credit paper, a policy they are about to issue. The day runs on that object. A sample email about a generic topic wastes people who can already write an email, and it never reaches the decision they are paid to make.
There is a second reason to pick a real object. PwC describes itself as client zero: it used the tools on its own work, then took that experience to clients. The piece says the firm is actively engaged on generative AI with 950 of its top 1,000 US consulting clients. An executive team that asks the company to move, and has not tried the tool on one internal artefact, is asking for a standard it has not met. One task in the room closes that gap in public.
Name the stack as well. PwC lists alliances with AWS, Anthropic, Google, Meta, Microsoft and OpenAI, and with enterprise software that already has these features inside it. A leadership team in a company of any size is rarely on a single logo. The session should use the tools people are actually allowed to open, and it should practise what they will say when a new one appears on a laptop the following week.
What they should do in the room
Ask each person to bring one decision they own in the next quarter. A budget line, a client-facing use, a hiring case, a policy they will have to defend. The provider builds the exercises on those decisions, with dummy data or data your security team has cleared. A case invented for another industry fills the hour. The decision this group will sign is the one that belongs on the table.
Then run the task. A board pack, a risk note, a first pass at a proposal. The room compares the output with the standard they would accept from a colleague. They write the check in plain words: who reads it, what they look for, and what they do when the prose is smooth and a fact is wrong. PwC's line that people stay in control of high-value decisions means something only if the room practises that control on a document they recognise.
Close the exercise with the sentence they will use in the next staff meeting: what is allowed this month, what stays out, and who answers when the case is unclear. If that sentence cannot be said aloud, the line is not ready and the session will talk around it. Write the sentence, then practise saying it. Managers who run the team's week need their own session on the tasks the team repeats, which is the subject of what AI training for managers should change. What a single day should leave a working team able to do is in what a corporate AI training day should produce.
How to brief the provider
Collective Campus builds high-performance teams for the AI world. The AI training workshop is the room for this leadership session. Send a one-page brief before anyone drafts an agenda. Name the decisions. Name the roles in the room, and keep the room to people who share those decisions. Name the tools they may use and the data that stays out. Name what you want each person able to do on the following Monday, in one sentence each.
Ask three questions and keep the written answers. Which of our decisions will the exercises use? Where in the day does each person produce a piece of work a colleague could mark? What will we look at four weeks later to see that a meeting has changed? A module list and a completion certificate leave those questions unanswered. You are paying for a change in the way this group decides.
Keep the group small enough that every person runs the task. A leadership offsite of forty, with a demonstration at the front, recreates an all-hands in a nicer room. People who share a decision, a finance leadership team or one executive group, can argue a real case. A mixed room of every function drifts back to definitions, because nobody's artefact is on the table. Put a history of the technology, a tour of ten products, a debate about which lab will win the year, and a future-of-work slide with no document from this company on a reading list. Spend the paid hours on the decision, the artefact and the check.
What good looks like a month later
Four weeks on, ask for the decision they brought. Has the workflow around it changed, and who checks the output? Ask for the sentence they used with staff, and whether anyone has repeated it since. PwC can point at hours, participation and a productivity range because the firm measured use. You do not need that scale. You need one changed decision and a check with a name on it. If both are missing, the day was a briefing, and the next one should be rewritten before you book a second cohort.
Before you lock the brief
- Which decision will this group sign in the next quarter, in one sentence?
- Which artefact from their own week will they run in the room, and on what data?
- Who checks a high-value output, and what do they do when it is fluent and wrong?
- What sentence will they say to staff about what is allowed, and where is that sentence written?
Send the brief with those four answers filled in as far as you can. We will shape the day around the decision, the artefact and the check. The Monday signature is the test of whether the session worked.
